Integrating Machine Learning and Predictive Simulation: From Uncertainty Quantification to Digital Twins

Industrial design and engineering, as well as healthcare, are facing unprecedented challenges with product complexity and increasingly faster innovation cycles. On the one hand, advances in computer modeling and simulation have resulted in a dramatic increase in the complexity of problems that can be solved; however, these methods are mostly limited by the availability of sufficiently accurate simulation models. Though engineering models may lack flexibility, they can have very strong predictive capabilities if the underlying mechanisms are captured well.

On the other hand, machine learning methods, like neural networks and deep learning, offer a much broader opportunity to tackle even more challenging problems, where functional relations do not exist or are not well-understood. Machine learning has demonstrated the ability to deliver very valuable insights into complex problems, but without understanding the mechanisms, the predictive capabilities, especially for long-range extrapolations, can be questioned.

So far these two communities have addressed industrial design, engineering, and operation challenges separately. In this workshop we intend to explore how the mechanistic world of computations can be effectively married with the world of machine learning and data-driven decisions, creating a forum for experts and scientists from industry and academia. Among the topics the workshop will address are uncertainty quantification, surrogate-based optimization, and digital twins, which are “living” digital models that integrate simulation and machine learning with data and update and change as their physical counterparts change.

The goal is to spur innovation by creating new research collaborations involving researchers from industry users, industry tool providers, government, and academia. The content of the workshop will be inspired by real world problems from medtech, aerospace, oil and gas exploration, and industrial control systems.